declare

SKILL.md

Declare Onboarding Contract

Extract or define an interf.yaml onboarding contract declaring what you need from enterprise environments.

The contract is plain English. Each dependency is a what (what you need) and ready (how enterprise knows it's done).

Process

Step 1: Understand the Solution

  • Read README, package.json / pyproject.toml / go.mod, and key config files
  • Identify what this project does and who the enterprise users are
  • Understand what value it delivers to enterprise customers

Step 2: Extract Dependencies

Scan the codebase for things that depend on the enterprise environment. Think like an FDE: what would you need from the customer before you can go live?

System access

  • HTTP client calls, SDK imports, API references
  • Database connections, ORM configs
  • OAuth/SSO/SAML configurations
  • Cloud storage references (S3, GCS, Azure Blob)
  • Webhook configurations

Data requirements

  • What data access is needed?
  • Historical data for training or analysis?
  • Real-time data feeds?
  • Custom field mapping needed?

Human coordination

  • Who at the enterprise needs to be involved?
  • Data team for field mapping?
  • Security team for review?
  • Business owner for validation?
  • Executive sponsor?

Process requirements

  • Security review or assessment?
  • Legal/DPA review?
  • Change management for affected workflows?
  • End-user training?

Infrastructure

  • Test/staging environments needed?
  • Network access or firewall rules?
  • Compute resources?

Step 3: Write the Contract

Write interf.yaml to the project root:

name: project-name
version: 0.1.0
description: What it does, one line

requirements:
  - what: Plain English — what you need from enterprise
    ready: How enterprise verifies this is done — specific and testable

  - what: Another dependency
    ready: Another verification criteria

optional:
  - what: Something that improves the experience but isn't blocking
    ready: How to verify

Writing good what descriptions:

  • Write for a non-technical stakeholder when possible
  • Be specific: "Read/write access to your CRM contacts and opportunities" not "CRM access"
  • Include scope: "12 months of contact history" not "historical data"
  • Include effort estimates for human tasks: "~4 hours of your data team's time"

Writing good ready criteria:

  • Make it verifiable: "We can create a contact via API from our staging environment"
  • Include what success looks like: "Extraction accuracy above 95% on sample documents"
  • Be specific about who verifies: "Field mapping document signed off by both sides"

Step 4: Auto-map Canonical Types

For each dependency, check if it matches a known canonical type from the Agent Onboarding Protocol. If confident, add the canonical field. If unsure, skip it — the contract works without it.

Reference: load the protocol skill for the canonical type reference.

Step 5: Summarize

Tell the user:

  • Total dependencies found (needs vs optional)
  • Any dependencies that might be missing — common things FDEs forget:
    • Security review (almost always needed at enterprise)
    • Test environment (hard to validate without one)
    • Data team coordination (custom fields always need mapping)
    • Executive sponsor (rollouts stall without one)
  • Next steps: review the contract, then preview rollout with the preview skill
Weekly Installs
1
First Seen
3 days ago
Installed on
windsurf1
amp1
cline1
opencode1
cursor1
kimi-cli1